[olsr-dev] OLSR and consistency

[Bernd Petrovitsch]

On Thu, 2007-02-22 at 10:31 +0900, giuseppe de marco wrote:
> Thank you for your reply.
> I'll answer inline.

Yes, please!

> Bernd Petrovitsch wrote:
> > On Wed, 2007-02-21 at 09:04 +0900, giuseppe de marco wrote:
> >   
> >> Suppose we want to update the routing module (OLSR) with a modified 
> >> version of it, or also
> >> we want to change the parameters in all OLSR modules running inside the 
> >> nodes of an ad-hoc net.
> >>
> >> How we can handle it automatically?
> >>     
> >
> > - Deploy the new version
> > - update the config files: Changes in most variables are relatively uncritical
> >    - one can never rule out the possibility of (short-time) routing loops.
> >    The only critical thing is that "LinkQualityLevel" must be either "0" on all
> >    or no nodes (and "UseHysteresis" must have also the correct value since these
> >    are incompatible).
> >   Folks, please correct me if there is something important missing!
> > - restart the daemon 
> >   
> It's not totally true. Even if you use LQ but with different window size 
> among nodes, results are very difficult
> to predict or at least to understand. We tried in our testbed two simple 
> experiments:
> 1) all nodes with LQ and window size = 10 and then window size = 20
> 2) all nodes with LQ but different window size
> In 1) all is ok, and in particular with window size =20 throughput goes 
> better. In 2) sometimes the network runs, sometimes it doesn't.
> Same thing if we change the frequency of HELLO messages.

Hmm, do you have a place where you gather such experiences so that one
can read/find them?

[....]
> > FunkFeuer has AFAIK an unknown number of OLSRD versions (from the last
> > several years) in parallel (because it is a lot of work to update the
> > firmware on dozens of nodes so it is at most done if someone fixes a
> > problem there and the one is fed up with the old version) in the mesh
> > (with "LinkQualityLevel 2" - it is much better in larger nets) and it
> > works ATM.
> >
> >   
> I agree. But suppose the other side of the story. An enterprise, or a 
> mesh network ruled out by one authority. There are also other examples.

Yes, of course - that resembles (hopefully) on the administrative
responsibility.
I see the single node owner in a mesh net as one such"enterprise" (and
at least in Vienna there are several people with more than 1 node ....).

[...]
> > Personally I'm not fond of: updating software automatically on a net of
> > (more or less) independent nodes run/managed by different people.
> > <paranoid>I wouldn't allow something on my nodes (unless I'm in control
> > of it).</paranoid>
[...]   
> As the previous.
> This is another problem, I think it's related to security.

ACK.

> On the other hand, don't you allow your Linux/Windows machine to do 
> automated download/update of new softwares?

No, I start them always manually (and it took some time to get the f$§@#
%g Win-XP to shut up about the "dangerous" not activated automatic
update) - just to know which packages are updated (at least in den Linux
world) so if something goes wrong tomorrow I have a chance to know what
is different.

	Bernd
-- 
Firmix Software GmbH                   http://www.firmix.at/
mobil: +43 664 4416156                 fax: +43 1 7890849-55
          Embedded Linux Development and Services